CN102289671A - Method and device for extracting texture feature of image - Google Patents

Abstract

The invention discloses a method and a device for extracting a texture feature of an image. The texture feature of an original image can be extracted accurately and quickly and the robustness of the texture feature is improved. The method for extracting the texture feature of the image, which is provided by the embodiment of the invention, comprises the following steps: a grey level image of the original image is obtained, the grey level image is divided into a plurality of image blocks, and the gray level concurrence matrix of each image block is computed; the angular second moment, the contrast, the entropy and the relative features serve as the chosen texture feature and the texture feature of each image block is computed according to each gray level concurrence matrix; and the corresponding texture feature of the original image is obtained according to the texture feature of each image block.

Description

Extract the method and apparatus of image texture characteristic

Technical field

The present invention relates to the image graphics processing technology field, especially relate to a kind of method and apparatus that extracts image texture characteristic.

Background technology

Along with internet scale is increasing, text wherein, picture, multimedia messages become increasingly abundant.In order to find the Useful Information user need search for the data of flood tide.In search procedure, must extracting accurately, feature could position exactly.Textural characteristics is important feature for image, and textural characteristics and color, shape facility also are called three big features of image.

The method of texture feature extraction mainly contains three classes at present, and the first kind is based on the method for statistics, and main method just is to use gray level co-occurrence matrixes.Second class is based on the method for structure, generally have only when texture primitive greatly when enough being cut apart, just use this method.The 3rd class is based on the method for frequency spectrum, mainly is based on Fourier and wavelet transformation.The first kind is applicable to natural texture and artificial texture widely based on the method for statistics in these three class methods, is ripe, the most practical present method.

The gray level co-occurrence matrixes of image has reflected the integrated information of gradation of image about direction, adjacent spaces, amplitude of variation, and it is the local mode structure of analysis image and the basis of queueing discipline thereof.As the characteristic quantity of texture analysis, often not directly to use the gray level co-occurrence matrixes that calculates, but on the basis of gray level co-occurrence matrixes, calculate second degree statistics again, as the texture characteristic amount that extracts.

The inventor finds to exist at least in the prior art following defective in realizing process of the present invention:

The number of the second degree statistics that can calculate on the basis of gray level co-occurrence matrixes is more, for example, and angle second order distance, contrary variance, entropy and entropy, difference entropy and maximum correlation coefficient etc.Because calculated amount is too huge, all second degree statisticses that calculate gray level co-occurrence matrixes in use simultaneously are infeasible in practice.Common way is to choose the second degree statistics of some gray level co-occurrence matrixes as textural characteristics.Yet, in the prior art, the extraction scheme of the textural characteristics of a cover comparative maturity not being proposed also, choosing of textural characteristics is more any, can't locate exactly, and the extracting method of textural characteristics is also complicated, is left to be desired.

Summary of the invention

The embodiment of the invention provides a kind of method and apparatus that extracts image texture characteristic, can extract the textural characteristics that obtains original image quickly and accurately, improves the robustness of textural characteristics.

For achieving the above object, the technical scheme of the embodiment of the invention is achieved in that

The embodiment of the invention provides a kind of method of extracting image texture characteristic, and described method comprises:

Obtain the gray level image of original image;

Described gray level image is divided into a plurality of image blocks, and calculates the gray level co-occurrence matrixes of each image block;

Angle second order distance, contrast, entropy and correlated characteristic as selected textural characteristics, and according to described each gray level co-occurrence matrixes, are calculated the textural characteristics of each image block;

According to the textural characteristics of described each image block, obtain the pairing textural characteristics of original image.

The embodiment of the invention also provides a kind of device that extracts image texture characteristic, and described device comprises:

The gray level image acquiring unit is used to obtain the gray level image of original image;

Textural characteristics is chosen the unit, is used for angle second order distance, contrast, entropy and correlated characteristic as selected textural characteristics;

The image block division unit is used for the gray level image that described gray level image acquiring unit obtains is divided into a plurality of image blocks, and calculates the gray level co-occurrence matrixes of each image block;

Image block textural characteristics computing unit is used for according to described each gray level co-occurrence matrixes, and that calculates each image block chooses the textural characteristics that the unit is determined by described textural characteristics;

Original image textural characteristics computing unit is used for the textural characteristics according to described each image block, obtains the pairing textural characteristics of original image.

By as seen above-mentioned, the invention provides the novel textural characteristics of a cover and choose scheme and texture feature extraction scheme, by with the suitable piecemeal of original image, obtain the textural characteristics of original image based on each image block, can extract the textural characteristics that obtains original image quickly and accurately, improve the robustness of textural characteristics.

In the scheme provided by the invention, angle second order distance, contrast, entropy and the correlated characteristic of gray level co-occurrence matrixes are chosen for the feature that is adopted, significantly reduced calculated amount required in the practical application, facts have proved that selected and textural characteristics that extract can be realized data search exactly, information matches and information location under several scenes.

Description of drawings

In order to be illustrated more clearly in the embodiment of the invention or technical scheme of the prior art, to do to introduce simply to the accompanying drawing of required use in embodiment or the description of the Prior Art below, apparently, accompanying drawing in describing below only is some embodiments of the present invention, for those of ordinary skills, under the prerequisite of not paying creative work, can also obtain other accompanying drawing according to these accompanying drawings.

A kind of method flow synoptic diagram that extracts image texture characteristic that Fig. 1 provides for the embodiment of the invention one;

A kind of method flow synoptic diagram that extracts image texture characteristic that Fig. 2 provides for the embodiment of the invention two;

A kind of apparatus structure synoptic diagram that extracts image texture characteristic that Fig. 3 provides for the embodiment of the invention three;

The secondary original image of Fig. 4 for being adopted in the experiment one of this programme;

The eight secondary original images of Fig. 5 to Figure 12 for being adopted in this programme experiment two.

Embodiment

Below in conjunction with accompanying drawing of the present invention, technical scheme of the present invention is clearly and completely described, obviously, described embodiment only is the present invention's part embodiment, rather than whole embodiment.Based on the embodiment among the present invention, the every other embodiment that those of ordinary skills are obtained under the prerequisite of not making creative work belongs to the scope of protection of the invention.

The embodiment of the invention one provides a kind of method of extracting image texture characteristic, and referring to Fig. 1, described method comprises:

11: the gray level image that obtains original image;

12: described gray level image is divided into a plurality of image blocks, and calculates the gray level co-occurrence matrixes of each image block;

13: with angle second order distance, contrast, entropy and correlated characteristic as selected textural characteristics;

14:, calculate the textural characteristics of each image block according to described each gray level co-occurrence matrixes;

15:, obtain the pairing textural characteristics of original image according to the textural characteristics of described each image block.

By as seen above-mentioned, the invention provides the novel textural characteristics of a cover and choose scheme and texture feature extraction scheme, by with the suitable piecemeal of original image, obtain the textural characteristics of original image based on each image block, can extract the textural characteristics that obtains original image quickly and easily.

In the scheme provided by the invention, angle second order distance, contrast, entropy and the correlated characteristic of gray level co-occurrence matrixes are chosen for the feature that is adopted, significantly reduced calculated amount required in the practical application, facts have proved that selected and textural characteristics that extract can be realized data search exactly, information matches and information location under several scenes.

Understand the technical program for the ease of clear, some technical essentials of gray level co-occurrence matrixes are described.

In three dimensions, two pixels of a certain distance of being separated by, they have identical gray level, perhaps have different gray levels, if can find out the statistical form of the joint distribution between such two pixels, incite somebody to action highly significant for image texture analyses.Based on this thought, propose space gray level co-occurrence matrixes (Spatial Gray Level Co-Occurrence Matrix), or be called gray level co-occurrence matrixes.The model of gray level co-occurrence matrixes is a kind of texture analysis method with popularity that proposes under the prerequisite of the texture information that the space distribution relation between each pixel has comprised image in the supposition image, has also derived on this basis such as gray scale one difference co-occurrence matrix analytic approach, gray scale one energy co-occurrence matrix analytic approach and gray scale one gradient co-occurrence matrix method.

Suppose that piece image f has N in the horizontal direction _xIndividual pixel has N in vertical direction _yIndividual composition, the gray level of each pixel is N to the maximum _RNote:

L _x＝{1，2，...，N _x}

L _y＝{1，2，...，N _y}

G＝{1，2，...，N _g}

Then can be interpreted as the image f that treats texture analysis from L _x* L _yTo the conversion of G, promptly to L _x* L _yIn each point, corresponding gray scale that belongs to f can be expressed as f:L _x* L _y→ G.Gray level co-occurrence matrixes is defined as the function of direction θ and spacing distance d, is designated as:

The capable j column element of representing matrix i, wherein (i, j) ∈ G * G, θ=0 °, 45 °, 90 °, 135 °.To different θ, matrix element is defined as follows:

In the following formula | { ... } | refer to cardinality of a set, D=(L _x, L _y) * (L _x, L _y).

All θ directions of expression gray level co-occurrence matrixes i capable j column element, adjacent spaces are to have one to get the i value in the pixel of d, and another phase adjacency pair of getting the j value is counted.Here, d can be taken as d=1,2,3,4,8 equivalences.

A kind of method of extracting image texture characteristic that the embodiment of the invention two is provided is specifically described below.Referring to Fig. 2, comprise following processing:

21: the image that the original image unified specification is turned to the equal resolution size.

For the search that accurately is implemented in data in one group of original image or the location of information, before texture extracts, the original image specification is turned to the picture of unified size, for example, have the image of equal resolution size:, thus the accuracy and the validity of the texture information that extracts guaranteed.

22: be converted to required gray level image.

Obtain the gray level image of original image in this step, can require this gray level image to satisfy predetermined gray level.For example, when original image is gray level image A, yet when this gray level image A does not satisfy predetermined gray level, with the grey level transition of gray level image A for predetermined gray level, to obtain required gray level image.When original image is coloured image B, coloured image B need be converted to gray level image according to predetermined gray level.

Number of greyscale levels is many more, and calculated amount is big more, and arithmetic speed is slow more; And number of greyscale levels is very few, then can influence the effect of institute's texture feature extraction, and the present invention is by test of many times, and choosing above-mentioned predetermined gray level is 64 grades, can reduce calculated amount as far as possible under the prerequisite that guarantees institute's texture feature extraction quality.

Can earlier original image be converted to gray level image when needing (if), again the gray level image specification be turned to the gray level image under the required gray level; Also can be in a treatment step original image be converted to the gray level image under the required gray level.

23: the calculating of image block and gray level co-occurrence matrixes.

Described gray level image is divided into 25 image blocks, and calculates the gray level co-occurrence matrixes of each image block.

Quantity to gray level image institute divided image piece should be suitable, the improper performance and the quality that can influence institute's texture feature extraction that piecemeal quantity is chosen, and through test of many times, it is 25 that this programme is chosen institute's divided image piece.

24: each image block is pressed the different directions texture feature extraction.

The textural characteristics that is extracted is selected in advance textural characteristics, because the calculated amount of gray level co-occurrence matrixes is very big, through experiment, selected textural characteristics comprises angle second order distance, contrast, entropy and relevant 4 features in this programme.

1) the angle second order is apart from (energy)

The angle second moment is the quadratic sum of each element of gray level co-occurrence matrixes, claims energy again.It is the tolerance of image texture grey scale change homogeneous, has reflected gradation of image be evenly distributed degree and texture fineness degree.If all values of co-occurrence matrix equates that all then the ASM value is just little; On the contrary, if big other value of some of them value is little, then the ASM value is big.One width of cloth has the gray level co-occurrence matrixes of consistent image gray to have only a value, and it equals the total pixel number of image, and its ASM value is maximum.Therefore, the value of ASM greatly then shows a kind of texture pattern than homogeneous and rule variation.The These characteristics of angle second moment become this programme choose its for the Consideration of textural characteristics it

Can utilize following formula in this programme, according to gray level co-occurrence matrixes calculated characteristics angle second order apart from ASM:

$\mathrm{ASM}=\underset{i=1}{\overset{L}{\mathrm{\Σ}}}\underset{j=1}{\overset{L}{\mathrm{\Σ}}}{\left\{p(i,j)\right\}}^2$

Wherein, (L represents gray level to p for i, the j) element of the capable j row of expression gray level co-occurrence matrixes i.

2) contrast (moment of inertia)

Contrast is near the moment of inertia the gray level co-occurrence matrixes principal diagonal, the value of its metric matrix be how to distribute and image in what of localized variation, reflected the sharpness of image and the rill depth of texture.The rill of texture is dark, and correlative value is big, and effect is clear; Otherwise correlative value is little, and then rill is shallow, and effect is fuzzy.The These characteristics of contrast becomes this programme and chooses it and be one of Consideration of textural characteristics.

Can utilize following formula in this programme, according to gray level co-occurrence matrixes calculated characteristics contrast C ON:

$\mathrm{CON}=\underset{i}{\mathrm{\Σ}}\underset{j}{\mathrm{\Σ}}{(i-j)}^{2}p(i,j)$

Wherein, p (i, j) element of the capable j row of expression gray level co-occurrence matrixes i.

3) entropy

The randomness of entropy tolerance image texture.When all values in the co-occurrence matrix of space all equated, it obtained maximal value; On the contrary, if when the value in the co-occurrence matrix is very inhomogeneous, its value is less.Therefore, the maximal value of entropy hint gradation of image distributes very at random.The These characteristics of entropy becomes this programme and chooses it and be one of Consideration of textural characteristics.

Can utilize following formula in this programme, according to gray level co-occurrence matrixes calculated characteristics entropy ENT:

$\mathrm{ENT}=-\underset{i}{\mathrm{\Σ}}\underset{j}{\mathrm{\Σ}}p(i,j)\log (i,j)$

Wherein, p (i, j) element of the capable j row of expression gray level co-occurrence matrixes i.

4) relevant

Correlation metric space gray level co-occurrence matrixes element be expert at or column direction on similarity degree, therefore, the correlation size has reflected local gray level correlativity in the image.When the matrix element value evenly equated, correlation was just big; On the contrary, if matrix pixel value differs greatly then correlation is little.If the horizontal direction texture is arranged in the image, then the COR of horizontal direction matrix is greater than the COR value of its complementary submatrix.Relevant These characteristics becomes this programme and chooses it and be one of Consideration of textural characteristics.

Can utilize following formula in this programme, according to the relevant COR of gray level co-occurrence matrixes calculated characteristics:

$\mathrm{COR}=\frac{1}{{\mathrm{\σ}}_x{\mathrm{\σ}}_y}\{\underset{i=1}{\overset{L}{\mathrm{\Σ}}}\underset{j=1}{\overset{L}{\mathrm{\Σ}}}\mathrm{ijp}(i,j)-{\mathrm{\μ}}_x{\mathrm{\μ}}_y\}$

Wherein, (L represents gray level to p, σ for i, the j) element of the capable j row of expression gray level co-occurrence matrixes i _x, σ _yThe expression variance, μ _x, μ _yThe expression average.

This programme is by the mode of choosing of above-mentioned textural characteristics, for the extraction of textural characteristics in the practical application provide a cover can with reference to and as the scheme of standard, in order to further specify the beneficial effect of this programme, as a comparison, the second degree statistics of other gray level co-occurrence matrixes that can calculate is carried out simple declaration, for example, variance f ₁, contrary variance f ₂, and average f ₃, and variance f ₄And f ₆, and entropy f ₅, difference entropy f ₇, relevant information estimates f ₈And f ₉And maximum correlation coefficient f ₁₀, specific as follows:

(1) variance $f1=\underset{i=1}{\overset{L}{\mathrm{\Σ}}}\underset{j=1}{\overset{L}{\mathrm{\Σ}}}{(i-\mathrm{\μ})}^{2}p(i,j)=\mathrm{\Σ}{(i,\mathrm{\μ})}^2{p}_x\left(i\right)$ μ is p (i, average j) in the formula;

(2) contrary variance $f2=\underset{i=1}{\overset{L}{\mathrm{\Σ}}}\underset{j=1}{\overset{L}{\mathrm{\Σ}}}\frac{1}{1+{(i-j)}^2}p(i,j)$

(3) and average $f3=\underset{i=2}{\overset{2L}{\mathrm{\Σ}}}{\mathrm{ip}}_{x+y}\left(i\right)$

(4) and variance $f4\underset{l=2}{\overset{2L}{\mathrm{\Σ}}}{(i-f_6)}^2{p}_{x+y}\left(i\right)$

(5) and entropy $f5=\underset{j=2}{\overset{2L}{\mathrm{\Σ}}}{p}_{x+y}\left(i\right)\log \left[p(i,j)\right]$

(6) and variance f6=p _X-y

(7) differ from entropy $f7=-\underset{i=2}{\overset{L}{\mathrm{\Σ}}}{p}_{x-y}\left(i\right)\log \left[p_{x-y}\left(i\right)\right]$

(8) relevant information is estimated $f8=\frac{\mathrm{HXY}-\mathrm{HXY}1}{\max (H_X-H_Y)}$

$f9={\{1-\exp [-2.0(\mathrm{HXY}2-\mathrm{HXY})\left]\right\}}^{\frac{1}{2}}$

H in the formula _xBe P _xEntropy, H _yBe P _yEntropy,

$\mathrm{HXY}=-\underset{i=1}{\overset{L}{\mathrm{\Σ}}}\underset{j=1}{\overset{L}{\mathrm{\Σ}}}p(i,j)\log \left[p(i,j)\right]$

$\mathrm{HXY}1=-\underset{i=1}{\overset{L}{\mathrm{\Σ}}}\underset{j=1}{\overset{L}{\mathrm{\Σ}}}p(i,j)\log \left[p_x\left(i\right)p_y\left(j\right)\right]$

$\mathrm{HXY}=-\underset{i=1}{\overset{L}{\mathrm{\Σ}}}\underset{j=1}{\overset{L}{\mathrm{\Σ}}}{p}_x\left(i\right)p_y\left(j\right)\log \left[p_x\left(i\right)p_y\left(j\right)\right]$

(10) the capable j column element of i of matrix Q in the second eigenvalue of maximum formula of maximum correlation coefficient f10=matrix Q: $Q(i,j)=\underset{k=1}{\overset{L}{\mathrm{\Σ}}}\frac{p(i,k)p(j-k)}{p_x\left(i\right)p_y\left(j\right)}$

This programme passes through test of many times, above-mentioned 9 features are all screened, and choose angle second order distance, contrast, entropy and relevant 4 features as the textural characteristics that is extracted, and can provide under the prerequisite of effective textural characteristics, greatly reduce the calculated amount of feature extraction.

Further, this programme according to 0 °, 45 °, 90 ° and 135 ° of 4 directions, calculates the textural characteristics of each image block according to the gray level co-occurrence matrixes of each image block.Promptly each image block is extracted 4 characteristic angle second orders, relevant, entropy, contrast by 4 directions respectively, extracted 16 textural characteristics of 4 directions altogether.

25: the textural characteristics that obtains original image with rotational invariance.

When the textural characteristics of described each image block comprises in different directions textural characteristics of each image block, in (as on) on each direction in 0 °, 45 °, 90 ° and 135 ° of 4 directions, calculate the average and the standard deviation of angle second order distance, contrast, entropy and the correlated characteristic of each image block, and with result of calculation as the pairing textural characteristics of original image.The average of all images piece that promptly just calculates and standard deviation are as the textural characteristics of original image.

Conversion may take place according to the textural characteristics that different directions calculated, this programme calculates variance and standard deviation to the same category feature of 4 directions of extraction, not only make the textural characteristics that obtains have rotational invariance, further improved the robustness of the textural characteristics that finally obtains; Also reduce the quantity of the textural characteristics that finally obtains, be convenient to carry out fast data search or information location.Under this mode, to having only 8 features, quantity has reduced half to each image block, then when being divided into 25 image blocks, whole original image has been obtained 25*8=200 textural characteristics from 16 features.

By as seen above-mentioned, the invention provides the novel textural characteristics of a cover and choose scheme and texture feature extraction scheme, by with the suitable piecemeal of original image, obtain the textural characteristics of original image based on each image block, can extract the textural characteristics that obtains original image quickly and accurately, improve the robustness of textural characteristics.

In the scheme provided by the invention, angle second order distance, contrast, entropy and the correlated characteristic of gray level co-occurrence matrixes are chosen for the feature that is adopted, significantly reduced calculated amount required in the practical application, facts have proved that selected and textural characteristics that extract can be realized data search exactly, information matches and information location under several scenes.

The embodiment of the invention three also provides a kind of device that extracts image texture characteristic.Existingly relatively obtain the scheme of textural characteristics by co-occurrence matrix, the outstanding advantage of this programme is to select effective 4 textural characteristics and original image to carry out suitable piecemeal, secondly is the textural characteristics that has obtained rotational invariance.Referring to Fig. 3, this device specifically comprises:

Gray level image acquiring unit 31 is used to obtain the gray level image of original image;

Textural characteristics is chosen unit 32, is used for angle second order distance, contrast, entropy and correlated characteristic as selected textural characteristics;

Image block division unit 33 is used for the gray level image that described gray level image acquiring unit 31 obtains is divided into a plurality of image blocks, and calculates the gray level co-occurrence matrixes of each image block;

Image block textural characteristics computing unit 34 is used for according to described each gray level co-occurrence matrixes, and that calculates each image block chooses the textural characteristics that unit 32 is determined by described textural characteristics;

Original image textural characteristics computing unit 35 is used for the textural characteristics according to described each image block, obtains the pairing textural characteristics of original image.

Further, described image block division unit 33 specifically is used for the gray level image that described gray level image acquiring unit 31 obtains is divided into 25 image blocks;

Described image block textural characteristics computing unit 34 specifically is used for according to described each gray level co-occurrence matrixes, according to 0 °, 45 °, 90 ° and 135 ° of 4 directions, calculates the textural characteristics of each image block;

Described original image textural characteristics computing unit 35 specifically is used in each direction, calculates the average and the standard deviation of angle second order distance, contrast, entropy and the correlated characteristic of each image block, and with result of calculation as the pairing textural characteristics of original image.

The concrete working method of each unit and module among apparatus of the present invention embodiment can be referring to the related content among the inventive method embodiment.

By as seen above-mentioned, the invention provides the novel textural characteristics of a cover and choose scheme and texture feature extraction scheme, by with the suitable piecemeal of original image, obtain the textural characteristics of original image based on each image block, can extract the textural characteristics that obtains original image quickly and accurately, improve the robustness of textural characteristics.

In the scheme provided by the invention, angle second order distance, contrast, entropy and the correlated characteristic of gray level co-occurrence matrixes are chosen for the feature that is adopted, significantly reduced calculated amount required in the practical application, facts have proved that selected and textural characteristics that extract can be realized data search exactly, information matches and information location under several scenes.

Below in conjunction with concrete experimental result, further specify the beneficial effect of this programme.

In experiment one, utilize this programme to extract image texture features.Referring to Fig. 4, a secondary original image that is adopted in the experiment one of this programme.Referring to table 1, shown an example utilizing this programme to extract image texture features among Fig. 4.

Table 1

Sequence number	Textural characteristics (texture)
		[0]	18.496535373405596
[1]	7.6545789154265442
		[2]	0.020827775524119272
[3]	0.007317312044616546
		[4]	25.130233006685344
[5]	0.53335814275488147
		[6]	0.0010293906204425249
[7]	0.000019704879619870455
		[8]	18.548006931596014
[9]	7.6600091029554767
		[10]	0.020230974101321197
[11]	0.0072160101311978886
		[12]	25.178929516339682
[13]	0.52043696287792252
		[14]	0.0010285893339953066
[15]	0.000020507881951226414

This programme is divided into 25 image blocks with the original image among Fig. 4 when texture feature extraction, shown the result who preceding two image blocks is extracted the textural characteristics that obtains in the table 1, and the precision of textural characteristics data is double precision (double).First image block corresponding sequence number 0 to 7, first image block corresponding sequence number 8 to 15 is followed successively by the variance and the standard deviation of contrast, angle second moment, entropy, correlativity by putting in order.As seen by above-mentioned, this programme can accurately carry out the extraction of textural characteristics.

The textural characteristics that utilizes this programme to extract in experiment two is classified to one group of original image.Referring to Fig. 5 to Figure 12, wherein, Fig. 5 to Fig. 7 has shown the image of animal skin, and the image of Fig. 8 and Fig. 9 grain of wood, Figure 10 are the image of rock texture, and Figure 11 and Figure 12 are the image of leather texture.When being applied to this programme in the above-mentioned 8 secondary original images, can each image texture features of rapid extraction, and carry out accurate classification.The image of Fig. 5 to Fig. 7 is divided into one group, the image of Fig. 8 to Figure 10 is divided into another group, the image of Figure 11 and Figure 12 is divided into another group, thereby realized that the image that will have similar textural characteristics accurately classifies.

The above only is preferred embodiment of the present invention, and is in order to restriction the present invention, within the spirit and principles in the present invention not all, any modification of being made, is equal to replacement, improvement etc., all should be included within the scope of protection of the invention.

Claims (10)

1. a method of extracting image texture characteristic is characterized in that, described method comprises:

Obtain the gray level image of original image;

Described gray level image is divided into a plurality of image blocks, and calculates the gray level co-occurrence matrixes of each image block;

Angle second order distance, contrast, entropy and correlated characteristic as selected textural characteristics, and according to described each gray level co-occurrence matrixes, are calculated the textural characteristics of each image block;

According to the textural characteristics of described each image block, obtain the pairing textural characteristics of original image.

2. method according to claim 1 is characterized in that, described according to described each gray level co-occurrence matrixes, the textural characteristics that calculates each image block specifically comprises:

According to described each gray level co-occurrence matrixes,, calculate the textural characteristics of each image block according to 0 °, 45 °, 90 ° and 135 ° of 4 directions.

3. method according to claim 1 is characterized in that, described textural characteristics according to described each image block obtains the pairing textural characteristics of original image and specifically comprises:

The textural characteristics of described each image block comprises each image block textural characteristics in different directions,

On each direction, calculate the average and the standard deviation of angle second order distance, contrast, entropy and the correlated characteristic of each image block, and with result of calculation as the pairing textural characteristics of original image.

4. method according to claim 1 is characterized in that, described described gray level image is divided into a plurality of image blocks, and the gray level co-occurrence matrixes that calculates each image block specifically comprises:

Described gray level image is divided into 25 image blocks, and calculates the gray level co-occurrence matrixes of each image block.

5. method according to claim 1 is characterized in that, the described gray level image that obtains original image specifically comprises:

Original image comprises gray level image or coloured image, according to predetermined gray level original image is converted to gray level image under the required gray level.

6. method according to claim 5 is characterized in that, described predetermined gray level is 64 grades.

7. according to each described method of claim 1 to 6, it is characterized in that before the described gray level image that obtains original image, described method also comprises:

The original image unified specification is turned to the image of equal resolution size.

8. according to each described method of claim 1 to 6, it is characterized in that,

Utilize following formula, according to gray level co-occurrence matrixes calculated characteristics angle second order apart from ASM, contrast C ON, entropy ENT and relevant COR:

$\mathrm{ASM}=\underset{i=1}{\overset{L}{\mathrm{\Σ}}}\underset{j=1}{\overset{L}{\mathrm{\Σ}}}{\left\{p(i,j)\right\}}^2$

$\mathrm{CON}=\underset{i}{\mathrm{\Σ}}\underset{j}{\mathrm{\Σ}}{(i-j)}^{2}p(i,j)$

$\mathrm{ENT}=-\underset{i}{\mathrm{\Σ}}\underset{j}{\mathrm{\Σ}}p(i,j)\log (i,j)$

$\mathrm{COR}=\frac{1}{{\mathrm{\σ}}_x{\mathrm{\σ}}_y}\{\underset{i=1}{\overset{L}{\mathrm{\Σ}}}\underset{j=1}{\overset{L}{\mathrm{\Σ}}}\mathrm{ijp}(i,j)-{\mathrm{\μ}}_x{\mathrm{\μ}}_y\}$

Wherein, (L represents gray level to p, σ for i, the j) element of the capable j row of expression gray level co-occurrence matrixes i _x, σ _yThe expression variance, μ _x, μ _yThe expression average.

9. a device that extracts image texture characteristic is characterized in that, described device comprises:

The gray level image acquiring unit is used to obtain the gray level image of original image;

Textural characteristics is chosen the unit, is used for angle second order distance, contrast, entropy and correlated characteristic as selected textural characteristics;

The image block division unit is used for the gray level image that described gray level image acquiring unit obtains is divided into a plurality of image blocks, and calculates the gray level co-occurrence matrixes of each image block;

Image block textural characteristics computing unit is used for according to described each gray level co-occurrence matrixes, and that calculates each image block chooses the textural characteristics that the unit is determined by described textural characteristics;

Original image textural characteristics computing unit is used for the textural characteristics according to described each image block, obtains the pairing textural characteristics of original image.

10. device according to claim 9 is characterized in that,

Described image block division unit specifically is used for the gray level image that described gray level image acquiring unit obtains is divided into 25 image blocks;

Described image block textural characteristics computing unit specifically is used for according to described each gray level co-occurrence matrixes, according to 0 °, 45 °, 90 ° and 135 ° of 4 directions, calculates the textural characteristics of each image block;

Described original image textural characteristics computing unit specifically is used in each direction, calculates the average and the standard deviation of angle second order distance, contrast, entropy and the correlated characteristic of each image block, and with result of calculation as the pairing textural characteristics of original image.

Images